Changes in customer searching behavior and trends they bring

The digital landscape is not just expanding as new products and services appear but is also rapidly evolving. Artificial intelligence, automation, digital transformation, and movements toward personalization are just a few factors impacting product development, especially in the E-commerce sector.

When it comes to internal search, the way business owners integrate it into their web platforms is also changing. There are lots of technological advancements and new approaches for internal search performance tuning. Besides, customer searching behavior is shifting, so in order to produce relevant internal search functionality, it is important to take such changes into account.

Let's take a closer look at all the changes in this domain so that you will be able to approach site search optimization with more accuracy.

Customer searching behavior is changing

Internal search is one of the key points of interaction for the customer with an online service. This website component takes into account what the customers want and acts as a router to quickly connect them to the right content.

• According to Google Cloud research, more than three-quarters (78%) of customers use the search function on retail websites. Site search is used more often than the navigation menu (49%), the filter feature (37%), and the homepage recommendations (30%).
• The research report by Coveo stated that 68% of customers are most likely to leave a brand’s website without purchasing because they simply can’t find what they are looking for.

At Wise Engineering, we believe exploring how customers actually search via your resource is one of the crucial parts of the site search development strategy. For the last few years, there have been multiple changes in customer searching behavior.

How people search now

The main shift is that search queries are getting more complex and detailed. Nowadays, customers tend to use more specific language to immediately indicate what they are looking for. Since the amount of information and things offered on web platforms is growing, this pattern helps customers navigate the vast digital libraries and find the needed items faster.

A lot of information → detailed search → the customers get what they want faster.

Customers specify one main keyword (topic) and add attributes (modifiers) that describe the keyword and add to it the necessary level of context.

Modifiers can range widely depending on the type of product or service being addressed. Customers use modifiers during the research and comparison phases of their journeys to establish a starting point for their purchases. They express what people want to know about a specific product (color, size, location, for example).

Modifiers are a good source to spot the early intent of customers. They provide clarity and direction on what is happening right now. As a result, you can look at people’s intent and, crucially, spot this early. Instead of trying to appeal to everyone with the same product, you can tailor your offerings to meet your customers' specific requirements.

Why is it important to spot early customer intent?

We are living in a period of economic uncertainty and witnessing soaring inflation that weighs on purchasing power. In Poland, for example, inflation surpassed 15% this summer and is expected to keep rising into 2023. Consumer confidence is decreasing, while the time required to conduct research is increasing. That is, users do a lot of research and then make a purchase.

How to spot early customer intent?

• We recommend using the Insights page in Google Ads. It shows trending search terms related to the keyword, so you can see new demand as it emerges. As a result, it allows you to keep in touch with your customers and change your advertising accordingly.

• You can also use Google Trends, which allows you to look at search interest for a topic or product category. It lets you see how search interest changes over time, which is useful because it may take time for intent to turn into demand.

When viewing search interest for a topic or product category in Google Trends, it may appear that growth has frozen and that there is no additional demand. When you become more specific and examine the changing modifiers individuals use, you can see their altering interests and desires.

Trends for internal site search in 2023-2025

Changes in customer search behavior emphasize the importance of spotting early customer intent and generally making internal search more convenient. This movement leads to the adoption of new technologies and practices for site search configuration. Here is a list of these techniques and some advice on how to enhance your internal search functionality.

#1 Query refinements and search result diversity

The first practice that we foresee as a growing trend in site search optimization is the usage of query refinements and search result diversity. The thing is, search queries may vary, and even if customers use more specific language, the intent can still be unclear. For instance, a search query may often be ambiguous and broad.

• Ambiguous queries may refer to several categories at once; for example, “mixer” may refer to a kitchen appliance or an audio component.
• Broad queries refer to one category, but there can be a variety of options for it, for example, shoes. This search query is unambiguous but under-specified.

Both search requests are complex, and internal search may simply not be designed to take them into account and process them correctly. The use of sophisticated approaches such as query refinement can be the right way to address such issues and narrow down search results. However, there is a slight but very important difference between processing ambiguous and broad queries.

• Processing ambiguous queries. From our experience, the best way to process ambiguous queries is to use a clarification dialogue to establish the searcher’s intent. In such a pop-up refinement, customers can specify a more detailed reference point. The pop-up window can contain several categories to which the request belongs. Then, the search query becomes more accurate, so the internal search can robustly map to an unambiguous intent.
• Processing broad queries. For this type of search query, we recommend using query refinement to narrow the results. Then, provide your customers with search result diversity – results containing various brands and models with different parameters and other technical characteristics. It allows customers to see a diverse selection of relevant results that showcase the variety across a few aspects (but not too many).

Search result diversity is not good for ambiguous queries, since a result set should not try to hedge between unclear, mutually exclusive intents. At the same time, for broad queries, it is important to strike a balance between search result diversity and search result desirability. If using only search result diversity, we may get a completely random distribution of styles and colors, which would be unlikely to serve the majority of searchers well. As a result, it is also critical to consider search result desirability.

To achieve this, we suggest using a greedy algorithm. It starts with the original order of results ranked by desirability, and then sequentially reranks them to improve diversity by reducing the divergence of the top-ranked results from the target distribution.

#2 AI for better query understanding

Another trend in site search optimization also relates to making the processing of search requests more intelligent. For the last decade, AI, more precisely word embeddings and deep learning, have shown great results in enhancing query understanding. You can train an AI-powered query categorization model that tracks all clicks or purchase data, as well as recognize equivalent queries from surface query similarity and post-search data.

Query understanding is the process of inferring the intent of the search query from the searcher’s keyword. Query understanding is useful not only for retrieval and ranking but also for interface decisions, recommendations, promotions, analytics, etc. One of the techniques for query understanding is query categorization, which also improves relevance and navigation. Here’s how AI can be used to enhance query categorization.

• Labeled training data from clicks or purchases. For example, the AI algorithm can track what search query users enter and then what clicks or purchases they make, so based on this behavior, it forms a product category. For this, any text classification model can be used, such as fastText and BERT.
• Apply the model to other queries. Compute categories for medium- to low-frequency queries online.

It is good practice to train query categorization using the categories of clicked and purchased products. Recognizing query equivalence sets the standard way to show what a search is about. When you can identify queries with the same (or nearly the same) intent, you can map poor-performing queries to better-performing equivalents. Other than that, you can intelligently recover from queries that return few or no results.

#3 Autocomplete and automatic rewriting

Autocomplete, rewriting, and typo-tolerance techniques can be another simple but smart solution for site search integration. All because, aside from using their browsing and search histories, these are additional ways to elevate customer searches. Using these techniques, you can guide customers to find the products they’re looking for even if they make mistakes when typing the keywords, or they are unsure what words to use to search for the products.

Autocomplete for improved searching experience

Autocomplete is more than just a way to help searchers type less. Other than saving time during searches, autocomplete has important advantages, such as better conveying the searcher's intent. When query understanding is working properly, autocomplete guides customers through the process of entering a search query. We believe that managing the trade-off between query probability and performance is the key to successful autocomplete.

Query probability. According to this technique, autocomplete predicts full search queries based on the partial ones that customers type. This process can be modeled using conditional probability. Given an incomplete question, you can offer searchers the most likely query that completes it. Query performance. We can consider clicks or actions (for example, purchases) as markers of search success. If clicks signify successful searches, you can count only the queries that result in clicks, rather than all the times a query occurs in the log. In other words, given the entered prefix, you select the autocomplete options that maximize the conditional chance of a click.

Automatic rewriting and typo tolerance techniques

Typo tolerance, as the name suggests, lets customers make mistakes when typing and still find what they need. Typo tolerance techniques acknowledge any mistakes customers make while entering keywords. Then, they allow for relevant results to be displayed when the words are close to an already existing phrase. We recommend the following typo-tolerance techniques.

• Clarification dialogue or "Did You Mean" suggestions. When internal search determines that a query is likely to be misspelled, it displays results for the original query but suggests an alternate query as a "did you mean" recommendation.
• Automatic rewriting. You may also enable site search to automatically rewrite the query and correct any spelling errors that customers may make. Typically, the interface offers a "search instead for" link that allows the searcher to bypass automatic rewriting and see results for the original query. Query rewriting can also include query expansion, segmentation, and scoping.

In all cases, the choice between a conservative "did you mean" and a more aggressive automatic rewriting with an opt-out should show how the internal search understands the query.

#4 Intuitive search

Another trend that owners of retail and web platforms should consider is making an internal site search more intuitive and convenient for customers. The main purpose of this trend is straightforward, namely to improve the accuracy of search results, as well as make it easier for users to navigate the site. Several factors contribute to the success of configuring intuitive internal search.

• Personalization

By personalizing the internal search, you can offer your customers a more tailored experience. You can use personalized search results to deliver personalized recommendations and offers to customers. Besides, it can also help you better understand customers’ needs and preferences. By analyzing the search patterns of customers, you can identify what products are most popular and adjust your offerings accordingly.

Among the key approaches for personalizing internal search are behavior-based ranking, rule-based ranking, personalized query probabilities, and more. You can also use ready-made tools for internal search configuration and personalization, such as Elasticsearch. It offers a wide range of possibilities for customizing search results right out of the box. Learn how to configure site search personalization from our previous post.

• UI and UX design best practices

UI and UX design tips might help you create an intuitive search. To start, place the search bar in a convenient location. Placing the search bar in an appropriate area, such as the top-right or top-center of the page, makes it noticeable and accessible. Every page should also contain a search box, so customers may search from anywhere on the site.

Another suggestion is to use a magnifying glass icon. This particular icon is a widely recognized sign of search and has universal user recognition, thus, we recommend you use it in your interface because it saves space and is recognized more quickly than words.

Adjust the size of your search bar. The size of your search bar should be determined by the relevance of the search bar on your site and the approximate length of a typical query. A common issue is making the input area too short, which makes it hard for customers to review and change their queries. As a general guideline, a 27-character text input will handle 90% of inquiries.

• Enhance filtering and product taxonomy

Finally, be sure to include filters and sorting tools to help customers narrow down search results and use relevant keywords for more precise searches. If your website's taxonomy is well-thought-out and puts into place, it will be easy for customers to find the information they need about a specific product.

If customers are searching for a black hooded puffer coat, for instance, they should have no trouble navigating to the coats section of your site owing to the intuitive organization of your taxonomy. They should be able to find the coat they want by using the search bar or one of the other main product filters, like the color filter.

#5 Natural language processing (NLP) to automate search

Natural language processing (NLP) aims to make searching a more user-friendly experience for customers. That’s why we foresee the adoption of NLP algorithms as another trend for site search optimization. These algorithms allow customers to enter more natural search queries, such as complete sentences or questions.

Normalization, typo tolerance, and entity recognition are examples of how NLP can make search more intelligent, especially semantic search. In contrast to lexical search, which simply looks for literal matches for query terms and returns results when a keyword is matched, semantic search recognizes the whole meaning of a query or the intent behind the words.

One of the possible ways to use NLP for internal search is Inbenta's Search module, which is driven by Symbolic AI. This tool can figure out what people are asking even if they use slang, jargon, or make spelling mistakes.

#6 Voice search

Another trend that is likely to impact internal site search optimization is the emergence of voice search. As voice assistants become more popular, customers will likely begin using them to search for products on websites. Analysts think that by 2024, 8.4 billion digital voice assistants will be in use around the world. This is because the market for voice assistants is growing quickly.

Voice search is similar to ordinary search queries, however, it is produced by the voice. Customers can enter search queries without using the keyboard. When customers make a voice request, the system translates the audio signal into text. The internal search will then process it and return the appropriate result. The following phases are involved in the processing of a voice request.

• Filtration – the targeted phrase is separated from the noise cloud. This is necessary to eliminate unwanted influence.
• Digitalization is the process of turning sound waves into computer-processable code.
• Analysis entails processing, compressing, and preparing the data for identification.
• Identifying Template Data – the query is compared to examples from the database and search history to identify specific words and phrases.

Voice search requires different algorithms and optimization techniques than traditional search, so online platforms and especially E-commerce must be prepared to adjust their search algorithms accordingly. It is currently available mainly to tech giants. However, utilizing speech recognition for a website will become increasingly easier in the future.

For instance, you can try Voxpow to use voice power quite easily and for free. It uses the best machine learning algorithms for speech recognition, reaching over 95% accuracy in over 100 languages.

Key takeaways

Paying attention to trends and how customers behave is not a whim, but a necessity to adapt to rapid change. Here are key takeaways from our article that will help you improve internal site search optimization in 2023-2025.

• Most of the customers have a specific need. When you are able to spot customer intent early, the chances of inferring relevant search results are much higher. We suggest analyzing the keywords and modifiers of the search requests and using the Insights page and Google Trends.
• Implement query refinement and search result diversity when appropriate. Search result diversity is useful for aspects that don’t represent constraints, but it still requires managing a trade-off with result desirability, as well as managing computational resources.
• AI-based search algorithms can help you better understand search queries. They can be used to search through large amounts of data to return the most relevant results. This may also help you to spot early customer intent.
• A more intuitive and accessible search experience can guide customers in formulating effective queries. This encouragement can come in the form of an intuitive search box, an improved autocomplete function or a natural-language interface such as one that uses voice search.

If you need guidance with setting up search functionality from scratch or optimizing your existing search engine, the Wise team is here to assist. We provide Elasticsearch consulting services that include end-to-end internal search configuration, performance tuning, query optimization, and more. Let's talk about what you want to accomplish with the internal site search and how your business goals can be achieved.